Thanks to Torgin Mackinga, Tim Daubenschütz, Christopher Goes and Justin Drake for comments.
In 2020, Daian et al. shook the cryptocurrency world by putting the spotlight on a glaring security and stability issue: Miner Extractable Value. In short: privileged actors have the ability to profit from other people's transactions. Sounds unfair right?
While solutions are being worked on to resolve the issue, MEV remains a captivating concept. The more places you look, the more you see similar issues appear. Privileged actors are able to exploit bottlenecks in domains ranging from politics to highway traffic to concert ticket sales.
Fortunately, we can do something about it.
In this post I take a look at value extraction across domains in order to gain an understanding of their respective solutions. But before we dive into the real world, let’s first build a model of value extraction.
What is value extraction exactly?
Value extraction is loosely used to mean unfair profit. It occurs in the context where multiple agents are trying to push transactions through a bottleneck. (UPDATE: some people are using the term in a broader sense: any advantage coming information assymetry).
What is fair, you ask? Well, you may think of the following (non-exhaustive) list of principles:
Fair ordering principle 1: time-based
Our first ordering principle is the oldest and most intuitive one. Time-based access is at first sight clearly democratic: if you want your transaction to be submitted, just try to be there before others. It works in many cases, but note that in some scenario's, it is impossible to perfectly define an order based on time when the bottleneck has multiple points of entry:
To highlight this using a simple example, consider three nodes, A, B, and C, that each receive 3 transactions, x, y, and z. A receives them in the order [x, y, z], B in the order [y, z, x] and C in the order [z, x, y]. Notice that a majority of nodes have received (x before y), (y before z) and (z before x)! This scenario, often called the Condorcet paradox, can cause a non-transitive global ordering even when all local orderings are transitive. - source
Fair ordering principle 2: auction-based
Any capitalist will quickly tell you that efficiently allocating resources on markets is achieved through auctions.
Auctions have been used for centuries to give access to a bottleneck to the highest bidder(s). If $X can be extracted from a transaction, any user should be willing to bid up to $X to extract the value. In an ideal world, this would lead to a maximum value transfer from agents to bottleneck operators, and potentially back to agents through redistribution and subsidies.
However, you can argue that this is no way to design a tax system. Running systems on auctions alone is dangerous. Auctions are only democratic in a perfect economic fantasy where every agent has equal power. However, as the auction amount has to be bid up front and the payoff may be uncertain, groups of transactions which are expensive, difficult and time consuming to extract value from have a higher risk to be influenced by wealthy agents.
Fair ordering principle 3: identity-based
As described above, wealth is still quite an indirect measure of individual capacity. We can also go straight to the heart of the matter and order transactions according to people's identity or personal characteristics. The simplest example would be to give each person a fixed maximum budget of transactions per period.
Of course, measurement is very difficult here, especially in private digital environments. Waste will occur in the form of agents trying to game the identification procedure. There are also endless philosophical debates possible about what an identity or group is exactly.
Identity-based access can also be combined with wealth based access, creating exciting mechanisms like quadratic voting/funding.
Fair ordering principle 4: lottery-based
In order to avoid unfair value extraction, batches of transactions can be ordered in a provably random way. Combined with time-based access, it both intuitively increases democratic access (no one can jump in line) but also decreases it (there is no way to signal increased commitment). However, a big problem is that it may encourage spam, enother form of wasting resources.
Fair ordering principle 5: Retrospective evaluation ordering
In many scenario's, you can measure agent's usage of the bottleneck, or even the impact that their transactions achieved. Ordering future transactions based on past impact ensures that the bottleneck is used by agents which make the best use of it. This requires identity-based access.
The problem with more complex measurements, is that they can be gamed and exploited. For example, in order to retain limited EU airport slots (bottleneck) airlines (agents) chose to deliberately waste resources by running empty flights...
Analyzing types of value extraction
Below I will have a look at three different domains where value extraction occurs in the real world.
Domain 1: Asset trading
Financial exchanges are prone to be manipulated by some highly suspect agents: themselves. The inner workings of traditional centralized exchanges are ultimately not transparant, which makes it possible for insider trading, price manipulation and many more scandals to occur.
On-chain decentralized financial exchanges (known as DEXes) enforce full transparancy, but therefore also risk value extraction by all involved parties. After a user submits a transaction to for example buy an asset, others may try to frontrun or sandwich it.
In a cryptocurrency context, block producers have an especially large advantage to order transactions any way they like and to access these arbitrage opportunities. The outsized power of block producers is seen as a threat to decentralization, inspiring proposer/block-builder separation and MEV auctions to give every user an equal chance to extract value. Moreover, by using cryptography, transactions can be encrypted and batch executed, minimizing the value that can be extracted from them.
Both centralized and decentralized exchanges do time-based processing of transactions, causing extremely large expenses towards colocation (more on this below). Valid transactions and blocks which are submitted first have precedence over those which are submitted later.
For fully fungible assets, identity-based ordering is not used in practice. In principle all transactions on a financial exchange, including the ones extracting value, could be carried out by only a few individuals. For non-fungible assets, such as tickets, collectibles, you name it, identity-based ordering does make sense, and I think there is a huge opportunitiy here to use this ordering principle more both in traditional and cryptocurrency contexts. You don't want your whole concert to be filled by a single person, and in many scenario's you don't want your collectibles to be owned by a single person either. And any intermediary institution which allows people to trade assets, runs the risk of unfair value extraction:
Domain 2: Messaging
Although on the internet information is often practically "free", given the zero marginal cost of replication, there are also many situations where information still has to pass through a centralized bottleneck. For information to reach a particular decisionmaking entity, it may have to compete with other information, thereby creating opportunities for extractable value.
Think for example of a whistleblower sending information to a political party about how the Dutch tax authority is systematically racist. Other agents may try to send their own proposals first, or to prevent the information from ever reaching the intended recipient. Another example is job candidates trying to submit their application for a hiring company. We see three main types of ordering:
- time-based ordering plays a role, but often both too much and too little. High level decisionmakers like politicians and CEOs are permanently overloaded. Therefore, on the one hand new information might be ignored because older vested interests make a more compeling case. On the other hand, new seemingly urgent emergencies appear every day, making it hard to focus on the important (long term) decisions.
- auction-based ordering can be used to allow information from the highest bidder to pass through. Paying politicians for speaking time is illegal, and it doesn't sound appealing to reach a big company's board, but it happens informally anyway through campaign financing, donations and personal favors. Formalizing this through auctions might not actually hurt, especially when combined with some form of limit per identity.
- identity-based ordering can be used to allow everyone a minimum level of information to share. The less financially privileged can and should be given frequent opportunities to voice themselves.
Domain 3: Traffic congestion
Our physical world is full of agents trying to pass through bottlenecks: cyclists on the bike path, cars on the highway or even airplanes passing through particular routes in the sky.
- time-based ordering is the norm, and value extraction takes a very particular shape in these scenario's. When congestion happens, agents may try to pass through earlier and earlier. Traffic jams start to occur, and agents may be waiting for a long time until they all finally passed through the bottleneck.
- auction-based ordering is occasionally used for high value transport like airplanes. Taxes for car drivers are becoming more and more common, and are a useful principle for more equitable usage of the roads.
- identity-based ordering is used, for example by governments to ensure only cars with particular number plates may drive on a particular day. This is quite an extreme measure, and may cause large economic misallocation to occur. It would be better to combine auctions with identity: power users are charged more than regular users.
Should profiting be minimized or maximized?
The term "value extraction" sounds negative and brings up dark imagery of exploitation and oil fields. Does that mean we should try to minimize value extraction?
Well, to use the other definition I mention above, we should minimize unfair profiting but absolutely maximize fair profiting. Arbitrage is fundamental to our modern economy. As Philip Daian described, there are many cryptocurrency systems which wouldn't survive without it:
- uniswap -> bribe anyone to bring price to market
- maker -> bribe anyone to pay gas to maintain stability
- 0x -> bribe anyone to provide orderbook liquidity
- CryptoKitties -> bribe anyone to pay gas for birthing
However, even when arbitrage opportunities are democratically accessible, another big risk arises: waste through true zero sum competition. The extreme investments into high frequency trading systems are wasting large amounts of resources through colocation and building hyperoptimized systems. The same can happen, and is happening, for many systems where anyone can profit and "extract value".
Fortunately, solutions to reduce zero sum competition exist (if you're curious, check out the link above!). We should aim to build systems which minimise the total societal cost of profiting.
Drivers of extractable value
There are multiple reasons why unfair profiting won't be fixed from one day to another.
First: fairness is relative. There is no single absolute fair ordering principles. This follows from the idea that we don’t have an absolute definition of fairness. Fairness is relative and political. See also this great blog for more commentary.
Second: power is unevenly distributed. Markets function best when power is distributed in an egalitarian way. We have a long way to go to achieve that goal (or perhaps we should try to go back in time).
Third: cryptography is eating the world. As a result, digital actions are more and more provable to the outside world. This ensures that agents can coordinate better, which sounds great, but it also ensures they can collude together. In a world of publicly verifiability (also called transcriptability), actions are susceptible to bribes.
A vision for fair value extraction
A lot of paradoxes later, the question remains, how do we prevent privileged actors from unfairly profiting and extracting value from our daily lives?
Perhaps it is best to draw lessons from taxation, which also grapples with complex social questions balancing fairness and efficiency. One of the most reliable rules to come forth from taxation is to: Broaden the base, lower the rate and to beware of distortionary effects on behavior. We should thus combine various methods, as noted also by Vitalik in his blog post about token sales.
In our not so distant past, we would have had to rely on social control and reputation to enforce these ordering principles. More recently, the world has been taken by storm by digital algorithms like Google Maps to coordinate in a better (and often lazy) way.
However, with the onset of modern cryptography, any group of agents can self-organize and run verifiable computation procedures to guarantee to others that they are supporting some notion of fairness. Call it emergent bottom-up opt-in private law. That is an exciting vision for the future.
The coordination arms race has begun.